Reproduce “AutoVRL: A High Fidelity Autonomous Ground Vehicle Simulator for Sim-to-Real Deep Reinforcement Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shathushan Sivashangaran, Apoorva Khairnar, Azim Eskandarian
AutoVRL, an open-source high-fidelity simulator built on PyBullet, OpenAI Gym, and Stable Baselines3, enables cost-effective and safe Deep Reinforcement Learning (DRL) for autonomous ground vehicles. It achieves successful sim-to-real transfer of DRL policies, allowing real-time deployment on physical hardware with low computational overhead.
Key insights
- A modular simulator architecture leveraging established open-source tools (PyBullet, OpenAI Gym, Stable Baselines3) effectively supports high-fidelity sim-to-real DRL for AGVs.
- Careful design of AGV digital twins, realistic sensor implementations, and environmental models in simulation is crucial for training policies that generalize to varied real-world conditions.
- While DRL policy training is computationally intensive, the real-time execution of trained policies is lightweight, making them viable for deployment on embedded robotic hardware.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.